Cursor — My Year in Code 2025

Source: Dev.to

Introduction

It’s been over a year since I switched from GitHub Copilot to Cursor. Reflecting on the past twelve months, I’ve seen heavy conversations among developers—both excitement and fear—as large language models (LLMs) have rapidly improved.

My Year with Cursor and LLMs

The release of Claude 3.5 was the first model that let me stay more hands‑off and still get decent results. Later, Composer arrived, followed by Claude 4.5, which at its core feels capable of replacing many junior engineers.

Seeing executives embrace AI while many engineers push back has been a common experience for anyone in tech or an affected industry.

The Emotional Landscape: Fear, Excitement, and Empowerment

• Fear: Hearing phrases like “our PM just built this tool for us” can feel unsettling, even if they’re partly satirical.
• Empowerment: Giving less‑technical people tools that let them shape products through rapid iteration is powerful. I remember learning PowerShell and Python to automate tasks on a locked‑down computer—fun and effective.
• Sharing: I loved showing teammates low‑code/no‑code tools, believing that once adopted they would improve both individual performance and the company as a whole.

Guardrails and Safety Concerns

The lack of guardrails worries me. Early tools like Quickbase offered sandboxed automation, keeping users safe while moving away from spreadsheets.

In contrast, the rapid rise of DeFi in 2020‑2021 introduced exciting possibilities but also massive security risks, leading many to lose money to scams. AI is following a similar pattern: new opportunities appear faster than the community’s ability to establish safety practices.

What the Developer Community Can Do

1. Make Learning Accessible: Continue creating AI‑focused educational content.
2. Teach Prompting: While prompting is often learned through experience, we can share best practices.
3. Promote Security: Develop resources that help non‑developers build applications safely.
4. Share Fundamentals: Emphasize basics of software engineering, UX, system design, and deployment.
5. Solve AI‑Era Problems:
   • Google’s SynthID for authenticity verification.
   • deadend.ai for detecting hallucinated links.

Looking Forward

• Use AI to augment workflows and build software that can change the world.
• Let AI handle time‑consuming tasks, freeing you to pursue ambitious ideas.
• Tools like Google’s Gemini can act as personal teaching assistants.
• Even if you struggle with web design, a friendly model can generate designs for you.

These tools don’t replace industry experts, but they give anyone a real chance to create impactful software.

---

Note: This isn’t an AI rant or a bullish piece; it’s an honest reflection on my personal journey—from fear to embracing AI, augmenting my work, and launching a startup. The opinions expressed are solely my own and do not represent my employers.